Hardware

Can HP Build the Computer of the Future?


Photonics research at HP Labs

Courtesy HP Labs

Photonics research at HP Labs

On June 11, Hewlett-Packard (HPQ) revealed plans to make a new kind of computer that it’s playfully calling The Machine. If HP can pull it off, it will mark a major rethinking of how computers are built. The design aims to combine huge advances in operating systems, memory, and data transfer technology to create a refrigerator-size computer able to store and analyze much of what an entire data center does today. Still years from the market, The Machine has become the talk of the industry, with rivals such as Dell mocking HP’s effort as “laughable” and other experts cheering on HP for trying something big and daring. “I think this is terrific,” says Greg Papadopoulos, a partner at venture capital firm New Enterprise Associates and a former computer architect at HP and Sun Microsystems. “This is new territory where people could get real benefits, and I hope they’re successful.”

It’s a funny thing to see technology zealots agog over the idea of a computer maker making a computer. Even just a decade ago, companies such as HP, Sun, IBM (IBM), and SGI (SGI) regularly announced plans to tackle bold frontiers in computing, and they had the talent on hand to meld departmental innovations into a working whole; they produced every major piece of an industrial-strength computer, from chips and storage to networking and operating systems. These days, such ambition is rare. “If you want to really rethink computing architecture, we’re the only game in town now,” says Martin Fink, HP’s chief technology officer and director of HP Labs, who has bet his career that The Machine will be a hit rather than a costly anachronism.

The story of the last 10 years of computing has been one of software and commodity parts. Companies such as Google (GOOG), Amazon.com (AMZN), Microsoft (MSFT), and Facebook (FB) needed so many servers to feed their consumer Web operations that they turned away from the specialized systems that had dominated computing for decades. Led by Google, they wrote software that made up for the shortcomings of cheap components. If a chip, hard drive, or stick of memory failed, it didn’t matter; automated software ordered another computer to pick up the work. These companies, which by and large give away their services for free, have reshaped data centers by designing them to perform calculations at the lowest possible cost.

“If you really want to rethink computing architecture, we’re the only game in town now”

It’s an approach that works fine—until the status quo changes. HP has designed its project to react to a moment it foresees about five years away, when it expects a new memory technology called the memristor to be ready for market. HP’s belief is that memristors will be much faster, more efficient, and able to store far more data than the high-speed DRAM and Flash memory used in servers today. To take full advantage of that power, a computer maker would have to revamp its operating system and data pathways to handle unprecedented volumes of information. It’s the kind of investment that consumer Web companies have largely avoided. “We have moved to what I call the Wall Streetization of technology,” says Shrijeet Mukherjee, the vice president for software engineering at Cumulus Networks, a maker of networking software. “It’s all about short-term gain.”

Mukherjee, who worked at SGI, says Web companies such as Google and Amazon have, in effect, achieved important advances for computer science with their high-efficiency data center software. They haven’t dedicated resources toward experimenting with radical advances in hardware. This is part of a generational shift, say Mukherjee and Fink, who both complain that few university students know how the guts of computers work. “There is a definite fear that we have stopped doing basic research around computer technology and that students are focused on much higher-level problems,” Mukherjee says. If HP’s hardware moonshot fails, he says he doubts a Facebook or Google will rise to take its place, but the R&D cash is there. “They certainly have the economic fortitude to make such a system,” he says. “It will require an individual who is willing to change the balance of things.”

The bottom line: The past decade’s shift of power from hardware to software companies has limited the development of computers.

Vance_190
Vance is a technology writer for Bloomberg Businessweek in Palo Alto, Calif. Follow him on Twitter @valleyhack.

Steve Ballmer, Power Forward
LIMITED-TIME OFFER SUBSCRIBE NOW

Companies Mentioned

  • HPQ
    (Hewlett-Packard Co)
    • $33.82 USD
    • -0.34
    • -1.01%
  • IBM
    (International Business Machines Corp)
    • $169.1 USD
    • -12.95
    • -7.66%
Market data is delayed at least 15 minutes.

Sponsored Links

Buy a link now!

 
blog comments powered by Disqus